Original article 99
Digit preference bias in the recording of emergency
department times
Thomas E. Locker and Suzanne M. Mason
Objective Digit preference bias has previously been
described in a number of different clinical settings. The
paper aimed to assess whether digit preference bias
affects the recording of the time patients arrive and
leave emergency departments.
Method An observational study of 137 emergency
departments in England and Wales was conducted. Each
department was asked to submit details of the time
of arrival and time of departure from the emergency
department for each patient attending during April 2004.
In addition, interviews with the lead clinician were
undertaken to determine the method used to record
the time of departure. The degree of digit preference
bias was assessed using a modification of Whipple’s
index.
Results One hundred and twenty-three (86.9%)
departments submitted data detailing 648 203 emergency
department episodes. 114875 (18.0%) episodes had a
recorded minute of departure of ‘0’ or ‘30’, with a further
2 81 890 (44.1%) having other values with a terminal digit
of ‘0’ or ‘5’. The mean modified Whipple’s index for time of
departure was 316.9 (range 70.9–484.4). Linear regression
demonstrates a small but significant inverse relationship
between the modified Whipple’s index and the mean total
Introduction
The total length of time patients spend in the emergency
department (ED) is a frequently used measure of
performance. Meaningful determination of this time is
dependent upon the accurate recording of both the time
the patients arrive in the ED and the time they leave.
Although it is commonplace for the patients’ time of
arrival in the ED to be recorded automatically by
computer as part of the registration process in the ED,
some EDs record the time patients leave the department
manually, this being transcribed from the ED records to a
database at a later stage. Manual systems of recording ED
times leave scope for digit preference bias to arise. Digit
preference bias describes the tendency for individuals to
record numbers that end with a particular digit such as
zero or five. Digit preference bias has been previously
described in situations when patients are asked to report
data such as age, year of menopause and smoking rate,
and in situations when clinicians are responsible for
recording measurements such as blood pressure and
birthweight [1–5]. The presence of digit preference bias
has also been used to detect fraud in clinical trial data [6].
time in department (b = – 0.05, 95% CIs – 0.09 to – 0.0004,
P = 0.048).
Conclusion Some departments show considerable digit
preference bias in the recording of time of departure
from the emergency department. Such bias may cause
difficulty in assessing changes in the performance of
departments.European Journal of Emergency Medicine
c 2006 Lippincott Williams & Wilkins.
13:99–101 European Journal of Emergency Medicine 2006, 13:99–101
Keywords: digit preference bias, emergency department
School of Health and Related Research, University of Sheffield, Sheffield, UK
Correspondence and requests for reprints to Thomas E. Locker, School of Health
and Related Research, University of Sheffield, Regent Court, 30 Regent Street,
Sheffield S1 4DA, UK
Tel: + 44 114 222 0868; fax: + 44 114 222 0749;
e-mail: t.locker@sheffield.ac.uk
Sponsorship: This study was conducted as part of a study funded by the National
Co-ordinating Centre for NHS Service Delivery and Organisation Research and
Development.
Conflict of interest: None.
Received 18 November 2005 Accepted 4 December 2005
The aim of this paper was therefore to determine
whether digit preference bias occurs in the recording of
times within EDs and whether this is influenced by the
use of manual vs. computerized systems to record this
information. A further aim was to explore whether there
was any association between the degree of digit
preference bias exhibited and the performance of the
EDs studied.
Methods
The ‘United Kingdom Waiting times in Accident and
Emergency Departments–Influence of Organizational
Factors Study’ (UWAIT) is a study of factors that
influence waiting time in EDs. All major EDs (type 1)
in England and Wales were invited to participate in the
UWAIT study. Major EDs are those that provide a
consultant-led 24 h service with full resuscitation facilities
and designated accommodation for the reception of
accident and emergency patients [7]. Each participating
department was asked to submit data for each new
patient attending their department during April 2004. For
each patient, the date and time of arrival and of leaving
c 2006 Lippincott Williams & Wilkins
0969-9546 Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
100 European Journal of Emergency Medicine 2006, Vol 13 No 2
the ED were collected. In addition, interviews were
conducted with the lead clinician in each department
to determine the methods used to record a patent’s time
of departure. The methods used to record the time of
departure were classified as ‘computerized’ if this was
undertaken using a computerized system at the time the
patient left the ED and ‘manual’ when the time was
recorded by hand in the patient’s ED notes and
transcribed to a computerized system at a later stage.
From these data, the minute of a patients’ time of arrival
and time of departure from the ED was determined for
each patient episode. The frequency of each value was
determined for each department. The degree of digit
preference bias for values ending in ‘0’ or ‘5’ was
determined using a modification of Whipple’s index.
The degree of bias was calculated according to the
following formula:
Modified Whipples index ðMWIÞ
P
ðn0 þ n5 þ n10 . . . n55 Þ
P
¼
100:
1=5 ðn1 þ n2 þ n3 . . . n59 Þ
This method assumes an even distribution of values from
‘0’ to ‘59’. The MWI may range from zero, when no values
end in ‘0’ or ‘5’, to 500 indicating that all recorded times
end in these values. A value of 100 indicates there is no
digit preference bias. The MWI was calculated for each
department.
The total time patients spent in the ED was defined as
the difference between the time of arrival and time of
departure, the mean of these times being determined for
each department. Linear regression was used to examine
the relationship between the mean total time in
department and the degree of digit preference bias,
assessed by the MWI.
Results
One hundred and thirty-seven departments consented to
take part in the UWAIT study. Of these, five (3.7%) were
unable to abstract the data from their information
technology systems and nine (6.6%) did not provide
data. One hundred and twenty-three (86.9%) departments submitted the required data detailing 648 203 ED
episodes. Of these, 9018 (1.4%) episodes had incomplete
data for either the time of arrival or time of departure and
were therefore excluded from further analysis.
Graph 1 shows the distribution of recorded minute of
arrival and departure. It can be seen that there is little
digit preference bias in the recording of time of arrival
with a mean MWI of 108.6 (range 66.1–279.4). The
recorded minute of departure shows considerable clustering around values with a terminal digit of ‘0’ or ‘5’.
114875 (18.0%) episodes had a recorded minute of
departure of ‘0’ or ‘30’, with a further 281 890 (44.1%)
having other values with a terminal digit of ‘0’ or ‘5’.
The mean MWI for time of departure was 316.9
(range 70.9–484.4).
The method of recording time of departure was known in
105 (85.4%) departments, of which 42 (40%) recorded
the time of departure using computerized systems and 63
(60%) recorded the time manually. Departments in the
latter group had a significantly higher mean MWI (375.7
vs. 213.7, t(103) = 6.312, P < 0.001).
Graph 2 shows the relationship between each department’s MWI and the mean total time patients spent in
Graph 1
10
Time of arrival
Time of departure
Percentage of episodes
9
8
7
6
5
4
3
2
1
0
0
5
10
15
20
25
30
35
40
45
50
55
Minute recorded
Distribution of the recorded minute of arrival and departure.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Digit preference bias Locker and Mason 101
Graph 2
Mean total tme in department (min)
450
Computerised
Manual
Unknown
400
350
300
250
200
150
100
50
0
0
100
200
300
MWI
400
500
600
Modified Whipple’s index vs. mean total time in the department by
method of recording time of departure.
that ED. Linear regression demonstrates a small but
significant inverse relationship between MWI and
the mean total time in department (b = – 0.05, 95% CIs
–0.09 to –0.0004, P = 0.048).
Discussion
This study shows that some EDs show marked digit
preference bias in the recording of the time that patients
leave the ED. This effect is more prominent in those
departments in which manual systems of recording time
are used. Does this matter? Limited evidence from this
study reveals that increasing degrees of digit preference
bias are associated with lower mean total times in the
department, although this effect is small. Why this effect
should occur is difficult to assess from the data collected.
It is possible that manual systems of recording the time
leaving the ED, although associated with increased digit
preference bias, might be quicker than recording the
times directly onto an information technology system.
Equally, it is possible that there is a tendency to
underestimate the time the patient leaves the ED when
this is recorded manually and hence those departments in
which such systems of recording are used exhibit greater
digit preference bias but show better performance.
minute of departure to be uniform from ‘0’ to ‘59’
minutes, then the expected proportion of times ending in
‘0’ or ‘30’ would be 3.3%. Thus, the data collected suggest
that 14.7% of episodes are having their time of departure
rounded up or down to the nearest 30 min. This may
cause difficulties when departments are attempting to
detect changes in their performance in response to
alterations in service provision. Performance targets for
EDs in England specify that 98% of patients should
spend no longer than 4 h in the ED. The data presented
in this paper cast doubt upon the ability of current
systems to accurately monitor the length of time patients
spend in the ED. This, in combination with the
distribution of total time patients spend in the ED that
we have previously reported, casts significant doubt upon
the reported performance of EDs in England [8].
Evidence shows that a change from manual to automated
systems of recording blood pressure results in lower
degrees of digit preference bias [4]. The finding that in
this study departments with computerized systems of
recording show less digit preference bias than those with
manual systems would suggest that a reduction in this
effect might be achieved with a move to computerized
systems of recording.
We would suggest that in future, part of the performance
monitoring of EDs should include an assessment of the
quality of data used to determine their performance.
Acknowledgement
The views expressed are those of the authors and not
necessarily those of the funding body.
References
1
2
3
4
5
6
Perhaps of greatest importance is the difficulty of
assessing changes in performance in departments that
exhibit large degrees of digit preference bias. In this
study, 18.0% of episodes had a recorded time of departure
ending in ‘0’ or ‘30’. If one assumes the distribution of the
7
8
Denic S, Khatib F, Saadi H. Quality of age data in patients from developing
countries. J Public Health (Oxford) 2004; 26:168–171.
Crawford SL, Johannes CB, Stellato RK. Assessment of digit preference in
self-reported year at menopause: choice of an appropriate reference
distribution. Am J Epidemiol 2002; 156:676–683.
Klesges RC, Debon M, Ray JW. Are self-reports of smoking rate biased?
Evidence from the Second National Health and Nutrition Examination Survey.
J Clin Epidemiol 1995; 48:1225–1133.
McManus RJ, Mant J, Hull MR, Hobbs FD. Does changing from mercury to
electronic blood pressure measurement influence recorded blood pressure?
An observational study. Br J Gen Pract 2003; 53:953–956.
Edouard L, Senthilselvan A. Observer error and birthweight: digit preference
in recording. Public Health 1997; 111:77–79.
Al-Marzouki S, Evans S, Marshall T, Roberts I. Are these data real? Statistical
methods for the detection of data fabrication in clinical trials. BMJ 2005;
331:267–270.
Report by the Comptroller and Auditor General. Improving emergency care in
England. London: National Audit Office; 2004. Report by the Comptroller and
Auditor General.
Locker TE, Mason SM. Analysis of the distribution of time that patients spend
in emergency departments. BMJ 2005; 330:1188–1189.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.